Mountain city land utilization change prediction method based on gray model and deep learning

Abstract

The invention discloses a mountain city land utilization change prediction method based on a gray model and deep learning, and the method comprises the following steps: S1, collecting data, and constructing a data set; s2, data preprocessing; s3, constructing a state transition matrix; s4, establishing a gray system model; s5, simulating and predicting the change trend of the land utilization type in combination with a Markov model; s6, training and evaluating the suitability probability of the land utilization type through a deep learning ANN module; s7, simulating the mutual relation among the land use types; s8, constructing a land utilization change prediction model according to the land utilization transfer change and the land utilization change driving factors; s9, performing multi-scene prediction on future land utilization change through the trained model; according to the method, efficient land utilization prediction is realized by integrating multi-source data, introducing a self-adaptive mechanism, combining a decision support tool and providing ecological environment influence assessment.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computer, in particular to a mountain city land use change prediction method based on grey model and deep learning. BACKGROUND

[0002] With the acceleration of urbanization, land use change has a profound impact on the ecological environment, economic development and social life. Compared with plain cities, the land use mode of mountain cities is particularly obvious due to the limitation of topography. Mountain cities are facing the problems of complex terrain, low land use efficiency and insufficient control of small-scale land use in urban fringe, so it is particularly important to rationally use land resources. Predicting land use change can help assess the potential impact of different projects and developments on the environment. By simulating future land use patterns, potential environmental risks and challenges can be better understood. However, under different development scenarios, the land use of mountain cities will show different development trends, so carrying out multi-scenario prediction of mountain city land use can help understand and master its change trend. Land use simulation and prediction models mainly include CA (Cellular Automata) model, FLUS (Future Land Use Simulation) model, Markov model and CLUE-S (Conversion of Land Use) model, etc. Using a single model to predict future land use change often has obvious limitations and cannot adapt to complex geographical environments, resulting in poor prediction results. SUMMARY

[0003] This section aims to summarize some aspects of the embodiments of the present application and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the abstract and title of the specification to avoid obscuring the purpose of this section, abstract and title, and such simplifications or omissions cannot be used to limit the scope of the present application.

[0004] In view of the above and / or existing problems in mountain city land use change prediction technology, the present application is proposed.

[0005] Therefore, the purpose of the present application is to provide a mountain city land use change prediction method based on grey model and deep learning, which combines the spatial analysis advantages of FLUS model and the time analysis advantages of grey Markov model to establish a new prediction model, so as to reduce the difficulty of formulating land conversion rules and reduce the excessive interference of human factors.

[0006] To solve the above technical problems, the present application provides the following technical scheme: a mountain city land use change prediction method based on grey model and deep learning, comprising the following steps:

[0007] S1, collect data and build a data set;

[0008] S2, data preprocessing;

[0009] S3, build a state transition matrix;

[0010] S4, establish a grey system model;

[0011] S5, combine Markov model to simulate and predict the change trend of land use type;

[0012] S6, through the ANN module of deep learning, train and evaluate the suitability probability of land use type, and simulate the mutual relationship between each land use type by using the CA module;

[0013] S7, according to the land use transfer change and the driving factor of land use change, a land use change prediction model is built;

[0014] S8, through the trained model, multiple scenarios of future land use change are predicted.

[0015] As a preferred scheme of the mountainous city land use change prediction method based on the grey model and deep learning in the application, wherein: in the step S1, the collected data includes land use image data of the prediction area and land use change driving factor data in four aspects of natural environment, social economy, accessibility and policy restriction.

[0016] As a preferred scheme of the mountainous city land use change prediction method based on the grey model and deep learning in the application, wherein: the step S2 specifically includes the following steps,

[0017] S201, according to the attributes of the data source, the land classification standard, the actual situation of the prediction area and the prediction target requirement, the classification standard of the land use type is determined;

[0018] S202, unify the data coordinate system of the grid data and convert it into the same projection coordinate system, trim the data range and overlap the research area, so that the number of all grid data is equal;

[0019] S203, according to the size of the prediction area, the model operation time and the calculation accuracy, the resolution of the grid data is determined.

[0020] As a preferred scheme of the mountainous city land use change prediction method based on the grey model and deep learning in the application, wherein: the step S3 specifically includes the following steps,

[0021] S301, the transition probability matrix between different land use types is calculated according to the historical data by using the Markov model, and the expression is,

[0022] (1);

[0023] (2);

[0024] wherein, p ij (m) represents the probability of transition from land class i to land class j after m time units, p ij is the probability of transition from type i to j from the beginning to the end, 0≤p ij ≤1, , S ij represents the area of type i to type j during the study period, represents the total area of land use type i, and n is the number of land use data categories.

[0025] As a preferred scheme of the mountain city land use change prediction method based on the grey model and deep learning in the application, wherein: step S4, specifically comprising the following steps,

[0026] S401, input data, construct a 7*5 matrix X with unit km 2 , the expression is,

[0027] (3);

[0028] 7 represents 7 types of land, 5 represents 5 years, 5 years for 1 year, x (i) (t) is the area of the i-th type of land in year t, i=1,…,7 respectively corresponding to paddy field, dry land, forest land, grassland, water area, construction land, unused land;

[0029] S402, data standardization, eliminate the difference of dimension,

[0030] (4);

[0031] z (i) (t) is the standardized area, the mean is 1;

[0032] S403, cumulative sequence generation,

[0033] (5);

[0034] wherein, z (1,i) (t k ) is the cumulative standardized area of the i-th type of land up to year ;

[0035] S404, establish a grey differential equation, predict the area of one type of land use, and the remaining 6 types are related variables;

[0036] (6);

[0037] Wherein, a is the development of gray number, reflecting the self attenuation rate; b i is the correlation weight of the i-th land use type to the paddy field change, z (1,1) is the standardized area of the first land use type, z (1,i) is the standardized area of the i-th land use type;

[0038] S405, model parameter estimation is carried out by least square method, parameter vector ,

[0039] (7);

[0040] (8);

[0041] S406, set sliding window, initial window: 1990-1995-2000 year data, indicate that the real data of 1995 and 2000 are used to predict the land use state in 2005; Slide to the next window: 1995-2000-2005 year data, predict the land use state in 2010, and so on, the time response function is obtained by accumulating the predicted value of the sequence ,

[0042]

[0043] Wherein is the standardized area of the i-th land use type at k time.

[0044] As a preferred scheme of the mountain city land use change prediction method based on grey model and deep learning in the application, wherein: the step S5, specifically includes the following steps,

[0045] S501, area residual calculation and parameter updating, fitting GM(1, N) for the data in the window, predicting the next time point, obtaining the area residual,

[0046] (9);

[0047] e (i) (t k+1 ) is the area residual of the i-th land use type in t k+1 year;

[0048] S502, parameter updating

[0049] (10);

[0050] (11);

[0051] wherein, η is a learning rate, controlling the update amplitude; e (1) is the prediction area residual of the model for the first type of land use data points, sign(e (1) ) is the area residual direction (positive / negative), determining the parameter increase / decrease, e (1) > 0, sign(e (1) ) = +1; e (1) = 0, sign(e (1) ) = 0, e (1) < 0, sign(e (1) ) = -1; a new is the updated development grey number, a is the development grey number before updating, b i,new is the updated correlation weight of the i-th type of land use, b i is the correlation weight of the i-th type of land use before updating;

[0052] S503, autoregressive correction based on area residual, fitting the historical area residual of each type of land use,

[0053] (12);

[0054] · is an autoregressive coefficient, estimated by least squares method; is the mean of the area residual; is the prediction value of the area residual of the i-th type of land at the next time (t k+1 ) in the time series;

[0055] S504, correcting the prediction value of the original sequence according to the current state and the transition probability matrix , the formula of which is,

[0056] (13).

[0057] S505, setting the transition probability between different land types according to different development scenarios, and combining the transition probability matrix and the corrected prediction value to calculate the future land use pixel number.

[0058] As a preferred scheme of the mountain city land use change prediction method based on the grey model and deep learning in the application, wherein, step S6 is specifically,

[0059] S601, selecting driving factors affecting land use change, normalizing the driving factor data using fuzzy membership degree, and obtaining a set of suitability probability maps of land use in the base year in the ANN module of the FLUS model, the expression of which is,

[0060] (12);

[0061] (13);

[0062] where i is land use type; s is hidden layer; r is grid; t is time; is suitability probability; is weight, ent s (r,t) is the feedback received by the grid r of the s-th hidden layer at the training time t;

[0063] S602, using the predicted year land use data and the constructed base year suitability probability map set, the CA module in the FLUS model is used to predict the land use spatial pattern in the predicted year, and the expression is, (14);

[0064] where TP t r,k is the total probability of converting to land use type p at the t-th iteration, represents the inertia coefficient of land use type k at time t, represents the spatial type conversion cost, represents the number of grids generated by land use type i after the iteration ends, N is the Moore neighborhood in CA, and w is the variable weight of each land use type;

[0065] S603, select the initial year of land use, land use change transfer matrix, land use transfer suitability image set and prediction period, carry out land use change simulation, and compare with the actual land use data, use Kappa coefficient to evaluate the accuracy of the simulation results, the calculation formula of Kappa coefficient is,

[0066] (15);

[0067] (16);

[0068] (17);

[0069] where n is the total number of grids; u1 is the number of simulated consistent grids; U is the number of land types; P0 represents the proportion of simulated consistent grids;

[0070] S604, error analysis is carried out on the simulation results, and the land use transfer rule is adjusted according to the analysis results, the model simulation is carried out again and the simulation results are evaluated until the simulation results meeting the accuracy requirements are obtained;

[0071] S605, analyze the prediction results, including the number of land use types and spatial transfer changes.

[0072] Compared with the prior art, the present application has the following technical effects: initial probability of the transition matrix is calculated by the Markov model, then the results of each land use type in the prediction year are corrected by the grey model, then the number of land use pixels in the future under different scenarios is calculated according to the corrected results and different land use transition probabilities, and finally the correction results are taken as the input quantity of the CA module, so that the prediction accuracy of the model is improved and the results are more in line with the actual land evolution process. BRIEF DESCRIPTION OF DRAWINGS

[0073] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor. Among them:

[0074] Figure 1 It is a flowchart of the present application.

[0075] Figure 2 It is a land use status image map of the present application.

[0076] Figure 3 It is a land use measured simulation comparison map of the present application.

[0077] Figure 4 It is a future land use change prediction map of the present application. DETAILED DESCRIPTION

[0078] In order to make the above-mentioned purposes, features and advantages of the present application more apparent and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the drawings of the specification.

[0079] In the following description, many specific details are set forth in order to provide a thorough understanding of the present application, but the present application can also be implemented in other ways different from those described herein, and those skilled in the art can make similar generalizations without departing from the connotation of the present application, therefore the present application is not limited by the specific embodiments disclosed below.

[0080] Secondly, the "one embodiment" or "embodiment" referred to herein means that the specific features, structures or characteristics can be included in at least one implementation of the present application. "In one embodiment" appearing in different places in the specification does not mean the same embodiment, nor is it an independent or alternative embodiment that excludes other embodiments.

[0081] Embodiment 1

[0082] Reference Figure 1 and Figure 4The embodiment of the application provides a mountainous city land use change prediction method based on a grey model and deep learning, comprising the following steps:

[0083] S1, collecting data and constructing a data set;

[0084] The collected data comprises land use image data of a prediction area and land use change driving factor data in four aspects of a natural environment, a social economy, accessibility and policy restrictions;

[0085] S2, data preprocessing, specifically,

[0086] S201, determining a land use type classification standard according to the attributes of a data source, a land classification standard, actual conditions of a prediction area and prediction target requirements;

[0087] S202, unifying raster data to a data coordinate system and converting the raster data into a same projection coordinate system, pruning data range and overlapping a research area, so that the number of all raster data is equal;

[0088] S203, determining the resolution of raster data according to the size of a prediction area, model operation time and calculation accuracy;

[0089] S3, constructing a state transition matrix, specifically,

[0090] S301, calculating a transition probability matrix between different land use types according to historical data by using a Markov model, and the expression is,

[0091] (1);

[0092] (2);

[0093] Wherein, p ij (m) represents the probability of transition from land class i to land class j after m time units, p ij is the probability of transition from type i to j at the initial and final stages, 0≤p ij ≤1, , S ij represents the area of transition from type i to j in the research period, represents the total area of land use type i, and n is the classification number of land use data;

[0094] S4, establishing a grey system model to predict the number of land pixels, specifically,

[0095] S401, inputting data and constructing a 7*5 matrix X with a unit of km 2 , and the expression is,

[0096] (3);

[0097] 7 represents 7 types of land, 5 represents 5 years, 5 years for 1 year, x (i) (t) is the area of the i-th type of land in year t, i = 1, …, 7 respectively corresponds to paddy field, dry land, forest land, grassland, water area, construction land, unused land;

[0098] S402, data standardization, eliminate dimensional difference,

[0099] (4);

[0100] z (i) (t) is the standardized area, the mean is 1;

[0101] S403, cumulative sequence generation,

[0102] (5);

[0103] Where, z (1,i) (t k ) is the cumulative standardized area of the i-th type of land as of year ;

[0104] S404, establish a gray differential equation, predict the area of one type of land use, and the remaining six types are associated variables;

[0105] (6);

[0106] Where, a is the development gray number, reflecting the self-decay rate; b i is the correlation weight of the i-th type of land to the change of paddy field, z (1,1) is the standardized area of the first type of land use, z (1,i) is the standardized area of the i-th type of land;

[0107] S405, model parameter estimation by least squares method, parameter vector ,

[0108] (7);

[0109] (8);

[0110] S406, set sliding window, initial window: 1990-1995-2000 data, represent the true data of 1995 and 2000, predict the land use state in 2005; Slide to the next window: 1995-2000-2005 data, predict the land use state in 2010, and so on, get the time response function by accumulating the generated sequence of predicted values ,

[0111]

[0112] wherein is the normalized area of the i-th land use type at time k.

[0113] S5, using the transition probability matrix obtained by the Markov model and the result of the grey model to predict the future land use state, specifically,

[0114] S501, area residual calculation and parameter updating, fitting GM(1, N) for the data in the window, predicting the next time point to obtain the area residual,

[0115] (9);

[0116] e (i) (t k+1 ) is the area residual of the i-th land use type at t k+1 year;

[0117] S502, parameter updating

[0118] (10);

[0119] (11);

[0120] wherein η is a learning rate, controlling the updating range; e (1) is the predicted area residual of the model for the data point of the first land use type, sign(e (1) ) is the direction of the area residual (positive / negative), deciding the parameter increase / decrease, e (1) > 0, sign(e (1) ) = +1; e (1) = 0, sign(e (1) ) = 0, e (1) < 0, sign(e (1) ) = -1; a new is the updated development grey number, a is the development grey number before updating, b i,new is the updated correlation weight of the i-th land use type, b i is the correlation weight of the i-th land use type before updating;

[0121] S503, autoregressive correction based on the area residual, fitting the historical area residual of each land use type,

[0122] (12);

[0123] • is the autoregressive coefficient, estimated by least squares; is the mean of the area residuals; is the predicted value of the area residual of the i-th land use type at the next time (t k+1 ) in the time series;

[0124] S504, according to the current state and the transition probability matrix, correct the predicted value of the original sequence , the formula is,

[0125] (13).

[0126] S505, according to the transition probability of different scenarios, adjust the transition probability matrix, so that the probability of each land use type is added to 1, and then calculate the number of each land use pixel in different scenarios according to the predicted value of each land use type after correction, as the input quantity of the CA module of the FLUS model.

[0127] S6, using deep learning algorithm, according to the land use transfer change and the driving factor of land use change, based on FLUS-Markov model, a land use change prediction model is constructed to predict the land use situation under different development scenarios in the future, and the mutual relationship between each land use type is simulated through the self-adaptive inertial cell mechanism;

[0128] Specifically,

[0129] S601, select the driving factor affecting the land use change, normalize the driving factor data by fuzzy membership degree, and obtain the suitability probability map set of the base year land use in the ANN module of the FLUS model, the expression is,

[0130] (12);

[0131] (13);

[0132] Wherein, i is the land use type; s is the hidden layer; r is the grid; t is the time; is the suitability probability; w s,i is the weight, ent s (r, t) is the feedback received by the grid r of the s-th hidden layer at the training time t;

[0133] S602, using the land use data of the prediction year and the constructed suitability probability map set of the base year, the CA (self-adaptive inertial cell mechanism) module in the FLUS model is used to predict the land use spatial pattern of the prediction year, the expression is,

[0134] (14);

[0135] wherein, TP t r,k is the total probability of the grid r being converted to land use type k at time t, represents the inertia coefficient of land use type k at time t, represents the spatial type conversion cost, represents the number of grids generated by land use type i after the end of iteration, N is the Moore neighborhood in CA, w i is the variable weight of the field of land use type i;

[0136] S603, select the initial year of land use, land use change transfer matrix, land use transfer suitability image set, pixel number of prediction year and prediction period, carry out land use change simulation, and compare with actual land use data, use Kappa coefficient to evaluate the accuracy of simulation results, the calculation formula of Kappa coefficient is,

[0137] (15);

[0138] (16);

[0139] (17);

[0140] wherein, n is the total number of grids; u1 is the number of simulation consistent grids; U is the number of land types; P0 represents the proportion of simulation consistent grids;

[0141] S604, error analysis is carried out on the simulation results, the land use transfer rule is adjusted according to the analysis results, the model simulation is carried out again and the simulation results are evaluated until the simulation results meeting the accuracy requirements are obtained;

[0142] S605, analyze the prediction results, including the number change and spatial transfer change of each type of land use.

[0143] S7, obtain the model input parameters meeting the accuracy through the deep learning algorithm, and then construct a land use change prediction model based on the FLUS-Markov model according to the land use data, the driving factors of land use change, the number of land use pixels in the prediction year and the suitability image set in the prediction year (it is a conventional technology, not the improvement point of the present application, and the specific implementation steps thereof are not described in detail in the present application);

[0144] S8, use the trained model to carry out multi-scenario prediction on future land use change.

[0145] The application calculates the initial probability of the transition matrix by the Markov model, then corrects the results of each land use type in the prediction year by the grey model, then calculates the number of land use pixels in the future under different scenarios according to the corrected results and different land use transition probabilities, and finally takes the correction results as the input quantity of the CA module, thereby improving the prediction accuracy of the model and making the results more consistent with the actual land evolution process.

[0146] Embodiment 2

[0147] As Figure 2 and Figure 3 As a second embodiment of the application, this embodiment verifies the technical effect of land use change prediction by using the application through simulation experiments.

[0148] This embodiment takes the simulation prediction of land use change in Chongqing as an actual application.

[0149] Regional profile:

[0150] Chongqing is located in the upper reaches of the Yangtze River in southwest China, with a total area of 82,400 km2, and currently governs 26 districts and 12 counties, of which mountains and hills account for about 98% of the total area, making it a typical mountain city. Chongqing is located in the north of Daba Mountain, the east of Wushan Mountain, the southeast of Wuling Mountain, and the south of Dalou Mountain. The main urban area is at an altitude of 168-400 meters, almost built on the mountains or slopes. However, rapid economic development and urbanization have led to increasingly prominent problems in land ecological environment and land use security pattern in Chongqing.

[0151] Land use image data of Chongqing from 1995 to 2015 (data resolution is 1 km) was collected, and data preprocessing (data cropping, correction, etc.) was performed by spatial analysis tools; land use classification standards were determined, and land use transition matrix was calculated; factors affecting land use transition were determined, and DEM, slope, aspect, GDP, population density, night light data, vegetation coverage, distance from river, distance from railway, distance from town, etc. were selected as driving factors for the construction of the model; land use transition suitability map set was made through ANN model; land use change prediction model was constructed based on FLUS-Markov model input, and land use simulation prediction was carried out.

[0152] In this embodiment, the land use types in Chongqing are classified according to the two-level type classification standard in Table 1, with cultivated land divided into paddy field and dry land, and other land use types divided into five categories according to the one-level type classification standard, a total of seven categories, based on which land use change simulation prediction is carried out. The land use transition suitability image made by ANN model is as follows Figure 2As shown in Figure 2, the land use change in 2015 simulated based on the land use change suitability atlas from 2005 to 2015 ( Figure 3 The Kappa coefficient of the comparison between the proposed method and the actual measurement reached 0.9092, which is higher than the accuracy of traditional CA (Cellular Automata) model, FLUS (Future Land Use Simulation) model, Markov model and CLUE-S (Conversion of Land Use) model (it is generally believed that a Kappa coefficient of 0.75 indicates a good simulation effect). This shows that the land use change prediction model constructed based on this method has high accuracy. At the same time, based on the land use pattern in 2015, the land use change pattern of Chongqing in 2025 under different development scenarios was predicted. Figure 4 shown.

[0153] Table 1 Land type classification standards

[0154]

[0155] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for predicting land use change in mountainous cities based on a grey model and deep learning, characterized by: The steps include: S1, collect data and build dataset; S2, data preprocessing; S3, constructing a state transfer matrix; S4. Establish grey system model; S5. Combine the Markov model to simulate and predict the changing trend of land use types; S6. Through the deep learning ANN module, the suitability probability of land use types is trained and evaluated, and the CA module is used to simulate the relationship between different land use types; S7. Construct a land use change prediction model based on land use transfer changes and land use change driving factors; S8. Use the trained model to make multi-scenario predictions of future land use changes.

2. The method for predicting land use changes in mountainous cities based on a grey model and deep learning as claimed in claim 1, wherein: In step S1, the collected data include land use image data of the prediction area and land use change driving factor data in four aspects: natural environment, social economy, accessibility and policy restrictions.

3. The method for predicting land use changes in mountainous cities based on a grey model and deep learning as claimed in claim 1, wherein: The step S2 specifically The following steps are included: S201. Determine the classification standard of land use type based on the attributes of the data source, the land classification standard, the actual situation of the prediction area, and the prediction target requirements; S202, unifying the data coordinate system of the raster data and converting it into the same projection coordinate system, trimming the data range to overlap with the study area so that the number of grid cells in all raster data is equal; S203: Determine the resolution of the grid data according to the size of the prediction area, the model operation time, and the calculation accuracy.

4. The method for predicting land use changes in mountainous cities based on a grey model and deep learning as claimed in claim 3, wherein: The step S3 specifically The following steps are included: S301. Use the Markov model to calculate the transition probability matrix between different land use types based on historical data. The expression is: （1）； （2）； Among them, p ij (m) represents the probability of moving from land type i to land type j after m time units, p ij is the probability of changing from type i to type j from the initial to the final period, 0≤p ij ≤1, , S ij represents the area that changes from type i to type j during the study period, represents the total area of ​​land use type i, and n is the number of classifications of land use data.

5. The method for predicting land use changes in mountainous cities based on a grey model and deep learning as claimed in claim 4, wherein: Step S4, specifically The following steps are included: S401. Input data and construct a 7×5 matrix X in km. 2 , its expression is, （3）； 7 represents 7 types of land, 5 represents 5 years, 5 years is 1 year, x (i) (t) is the area of ​​the i-th type of land in year t, where i = 1,…,7 corresponds to paddy field, dry land, forest land, grassland, water area, construction land, and unused land, respectively; S402, data standardization, eliminating dimensional differences, （4）； z (i) (t) is the standardized area with a mean of 1; S403, accumulate and generate a sequence, （5）； Among them, z (1,i) (t k ) is the land of category i as of the year The cumulative standardized area of S404. Establish a grey differential equation to predict the area of ​​one land use type, with the remaining six types as associated variables; （6）； Among them, a is the development gray number, reflecting its own decay rate; b i is the association weight of the i-th type of land to the change of paddy fields, z (1,1) is the standardized area of ​​the first land use type, z (1,i) is the standardized area of ​​land of type i; S405, use the least squares method to estimate the model parameters, the parameter vector , （7）； （8）； S406. Set a sliding window. The initial window is 1995-2000-2005 data, which means using the real data of 1995 and 2000 to predict the land use status in 2005; slide to the next window: 2000-2005-2010 data, and predict the land use status in 2010. And so on. The time response function is obtained by accumulating the prediction values ​​of the generated sequence. , ； in is the standardized area of ​​the i-th type of land at time k.

6. The method for predicting land use changes in mountainous cities based on a grey model and deep learning as claimed in claim 5, wherein: The step S5 specifically The following steps are included: S501, area residual calculation and parameter update, fit GM(1,N) to the data in the window, predict the next time point, and get the area residual. （9）； e (i) (t k+1 ) is the i-th type of land at t k+1 Area residuals for the year; S502: Parameter update （10）； （11）； Among them, η is the learning rate, which controls the update amplitude; e (1) is the residual error of the model’s prediction of the area of ​​the first land use type data point, sign(e (1) ) is the direction of area residual, which determines the increase or decrease of parameters, e (1) >0, sign(e (1) )=+1;e (1) =0, sign(e (1) )=0,e (1) <0, sign(e (1) )=-1; a new is the development gray number after updating, a is the development gray number before updating, b i,new is the updated association weight of the i-th land use type, b i is the associated weight of the i-th land use type before updating; S503, based on the autoregressive correction of area residuals, each type of land is fitted with its historical area residuals. （12）； · is the autoregressive coefficient, estimated by the least squares method; is the mean of area residuals; is the next moment (t k+1 ) predicted values ​​of area residuals; S504: Correct the predicted value of the original sequence according to the current state and the transition probability matrix , the formula is, （13）； S505. Set the transition probability between land types according to different development scenarios, and then calculate the number of pixels of future land use by combining the transition probability matrix and the revised prediction value.

7. The method for predicting land use changes in mountainous cities based on a grey model and deep learning according to claim 6, wherein: Step S6 specifically includes: S601. Select the driving factors that affect land use change, normalize the driving factor data using fuzzy membership, and obtain the suitability probability atlas of land use in the base year in the ANN module of the FLUS model. The expression is: （14）； （15）； Among them, i is the land use type; s is the hidden layer; r is the grid; t is the time; is the suitability probability; w j,i is the weight, ent s (r,t) is the feedback received by the grid r of the sth hidden layer at training time t; S602. Using the land use data of the forecast year and the suitability probability atlas of the base year, the CA module in the FLUS model predicts the land use spatial pattern of the forecast year. The expression is: （16）； Among them, TP t r,k is the total probability of converting to land use type r at the tth iteration, represents the inertia coefficient of land use type k at time t, Indicates the cost of spatial type conversion, represents the number of grids generated by land use type i after the iteration, N is the Moore neighborhood in CA, and w is the domain variable weight of each land use type; S603. Select the initial year of land use, land use change transfer matrix, land use transfer suitability image set, and forecast period, conduct land use change simulation, and compare with the actual land use data. Use Kappa coefficient to evaluate the accuracy of simulation results. The calculation formula of Kappa coefficient is: （17）； （18）； （19）； Where n is the total number of grids; u1 is the number of grids with consistent simulation; U is the number of land types; P0 represents the proportion of grids with consistent simulation; S604: Perform error analysis on the simulation results, adjust the land use transfer rules based on the analysis results, re-run the model simulation and evaluate the simulation results until simulation results that meet the accuracy requirements are obtained; S605. Analyze the forecast results, including the quantitative changes and spatial transfer changes of various land use types.

Images